Anthropic says any lab can now let a language model agent run the whole protein design stack

Anthropic has demonstrated that Claude models can autonomously manage specialized software tools to design proteins that bind to specific biological targets. By coordinating these external programs, the model achieved a 35 percent success rate in initial testing, notably higher than the typical industry average of 10 to 15 percent. This approach suggests that language models can serve as effective orchestrators for complex scientific workflows, potentially accelerating the early stages of drug discovery by automating technical processes that previously required extensive manual intervention.